Title:   Materials to reproduce analysis reported in: Elite-Public Interaction on Twitter: EU issue Expansion in the Campaign
Authors: Zoltan Fazekas, Sebastian A. Popa, Hermann Schmitt, Pablo Barbera, Yannis Theocharis
Date:    March 2020
Journal: European Journal of Political Research
---

Brief introduction:

Here you can find data and script to produce the analysis in the main text of "lite-Public Interaction on Twitter: EU issue Expansion in the Campaign". The data file is supplied for the sole purposes of reproducing the materials in the paper.

Raw Twitter data cannot be shared publicly, hence the text analysis and classifier training materials are not available here. Unique IDs have been added to merge back for any particular check requested. Politician user names have been removed in order to preserve anonymity, unique politician IDs are included. Please get in touch with the authors regarding data use for different purposes and original files, we are happy to help.

Contents:

1.  Data

ejpr-tweets.Rdata

	* .Rdata file one object, "cand_own" that is the tweet level dataset used for the main analysis, and includes candidate and party level features as well. Variables are described in the manuscript and in the "Variables" section at the end of this document.

cands-all.Rdata

	* .Rdata file one object, "cands_all" that contains country level social media use characteristics. Variables are described in the manuscript and in the "Variables" section at the end of this document.


2. Code

ejpr-engaging.R 

	* main analysis that includes model fitting, model based figures, and descriptives reported in the paper.

ejpr-fig-1.R: 
	
	* code to create Figure-1 in the manuscript.

Variables:
++++++++++
1. ejpr-tweets.Rdata

"tw_uid"           = unique tweet (observation) id.
"pol_uid"          = unique politician id.
"party"            = party of candidate.
"party_nest_id"    = generated party id.
"country"          = country.
"fc"               = candidate account follower count (+1 logged).
"fc_2sd"           = candidate account follower count (+ 1 logged), 
					 mean centered and divided by 2 SD (within country).
"ches_eu"          = Chapel Hill EU position for party of candidate (for appendix).
"ches_eu_2sd"      = Chapel Hill EU position for party of candidate (mean centered).  
"electability"     = 3 category electability variable.
"sitting"          = is candidate sitting MEP (1 = yes, 0 = no)
"eu"               = EU content, continuous (probability), 0 to 1 [based on classifier].
"engaging"         = style engaging, continuous (probability), 0 to 1 [based on classifier].
"is_eng"           = style, dichotomized engaging (EU content if > 0.5).
"is_eu"            = EU content, dichotomized eu (EU content if > 0.5).
"eu_mc"            = eu content mean centered (within politician).
"eng_mc"           = idem for engaging content.
"n_resp_indata"    = number of replies to the tweet by the politician.
"n_retweet_public" = number of retweets by accounts that are not politicians.

2. cands-all.Rdata

"Country"           = country.
"total_cands"       = total candidates in country.
"cands_on_twitter"  = candidates with a Twitter account.
"cands_act_twitter" = candidates active on Twitter (tweeted at least once in the period).
"total_tweets"      = total tweets (original tweets by candidates),
"med_tweets"        = median number of tweets.
"total_parties"     = total parties in country.
"total_responses"   = total number of replies. 
"med_responses"     = median number of replies.
"rho_twresp"        = correlation between tweets and replies (within country)
"rat_resptw"        = median mentions/tweets ratio.
"is_samp"           = is country in the content coded data (analysis), = 1 (for 4 countries).
"ac" 				= country abbreviation.

